In systems that handle a large volume of data, redundant arrays of inexpensive disks (RAID) apparatuses capable of achieving high reliability, and high read and write performance are used. A RAID apparatus is an apparatus that is redundantly constructed by connecting a plurality of storage devices (hereinafter referred to as disks), such as hard disk drives (HDD), or the like. In a system as described above, a network attached storage (NAS) apparatus, which is provided with a scheme that allows accesses from a plurality of terminal apparatuses through a network, and which allows central management of data, or the like is also used.
Even if a part of disks fails, a RAID apparatus is provided with a mechanism that allows restoring data that was stored in a failed disk from the data stored in the remaining disks. For example, a RAID apparatus stores parity data, and executes operation processing (hereinafter referred to as parity operation) based on the parity data so as to make it possible to restore the data stored in the failed disk. The restored data is stored in a replacement disk (hereinafter referred to as a spare disk) provided in the RAID apparatus. When data is stored in a spare disk, it is possible for a RAID apparatus to continue normal operation using the spare disk.
As described above, when a RAID apparatus is used, it is possible to continue normal operation even if a failure occurs in a part of the disks. However, if the amount of data to be restored is large, it takes a long time to restore the data by parity operation. That is to say, depending on the amount of data, and the use state, it sometimes takes a long time (hereinafter referred to as data reconstruction time) to perform processing (hereinafter referred to as restoration processing) to restore data that was lost by a disk failure.
In view of the circumstances described above, proposals have been made of a method for reducing data reconstruction time. For example, a proposal is made of a method in which data in an operating disk is stored in a spare disk in advance, and the data stored in the spare disk is used for restoration processing. Also, data stored in the other disks are copied to an unused area of each disk, and the copied data is used for restoration processing. Also, a proposal has been made of a method in which the data copied in a storage area (hereinafter, referred to as a spare area), which is provided in advance for each disk, is used for restoration processing.
Related techniques are disclosed in Japanese Laid-open Patent Publication Nos. 2003-108316, 2009-205571, and 2000-200157.
If a technique according to the above-described proposal (hereinafter, referred to as a proposed technique) is applied, it is possible to use data stored in a spare disk, an unused area, or a spare area (hereinafter referred simply as a spare area) in advance for restoration processing, and thus it is possible to shorten data reconstruction time compared with the case of executing the parity operation. However, the above-described proposed technique is based on the premise that all the data stored in each disk is copied to the spare area.
Providing a disk space capable of storing all the data stored in each disk as a spare area will result in an increase in cost of a RAID apparatus. On the other hand, if the capacity of the spare area is not sufficient, part of data might not be copied. The part of data that has not been copied is restored by parity operation in restoration processing, and thus it takes a long time for restoring the data.
The above-described proposed technique is based on the premise that all the data is copied to the spare area, and no consideration is given to the kind of data to be copied.
Accordingly, if the above-described proposed technique is applied, at the time of copy and restoration processing on the data, data to be restored in a short time, and the other data are handled in the same manner. That is to say, in the case where a spare area capable of storing all the data in each disk is not provided, there is a substantially same possibility of taking a long time for restoration regardless of the kind of the data.